Comment 4 for bug 1608446

Georgy, could you please enable logging in /etc/libvirt/libvirtd.conf (note the *d* suffix - this is the config for the daemon part of libvirt):

log_level = 1
log_outputs="3:syslog:libvirtd 1:file:/var/log/libvirt.log"

restart libvirtd:

sudo service libvirtd restart

and give it another try.

I can't reproduce it on my environment: all 10 booted instance on 2 compute nodes are always resumed properly after hard reset via virt-manager. This seems to be a libvirt issue at this point (nova only tells libvirtd to start a domain, libvirtd is the one that writes the apparmor profile to the file system).